M3Exam (Multimodal Memory) - Find-Matching Image: leaderboard
M3Exam (2026) is a multimodal conversational-memory benchmark of 5,150 questions over long multi-session user-assistant conversations with images and documents; the frontier models here answer with the gold supporting sessions in context. This board is exact match (0-100) on questions that ask which image file matches a description among hard negatives; higher is better.
Metric: Exact match (0-100) on the find-matching-image (FM) questions of 5,150 questions over 239 long multi-session user-assistant conversations with images and PDFs (15 personas); each model answers with the annotated gold supporting sessions and their images or rendered pages placed in context, no memory module: name the image file that matches a description among hard negatives; printed 0-1 and x100 here; higher is better. Source: arxiv.org. Saturation forecast: Around 2034. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 24.03 | #94 |
| 2 | Claude Opus 4.6 | 22.03 | #102 |
| 3 | Qwen 2.5 VL 7B Instruct | 20.7 | #1833 |
| 4 | Seed 2.0 Pro | 19.5 | #211 |
| 5 | GPT-5.4 | 18.9 | #123 |
| 6 | Grok 4 | 18.71 | #307 |
| 7 | GLM-5.1 | 16.67 | #179 |
| 8 | Qwen 3.6 Plus | 15.2 | #222 |
| 9 | Kimi K2.5 | 13.08 | #256 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=m3exam-multimodal-memory-find-matching-image · How It Works · Data refreshed daily, snapshot 2026-11-13.